Two ways to buy an AI build: pay by the hour and settle up as you go, or agree a fixed price per milestone. Neither is universally right — this guide shows which one carries the risk you actually want to carry.
Last updated: 11 June 2026
For most SME AI builds, fixed-price milestones are the safer choice: the supplier carries the overrun risk, you get budget certainty, and payment is tied to a working deliverable rather than hours logged. Time-and-materials (hourly) is genuinely better when the scope is open-ended research and nobody can honestly define 'done' yet. Crux Digits runs a fixed ladder — AI Audit & Strategy at €2,500, Proof of Concept at €20,000, production from €50,000 — and offers consulting at €150/hour for the open-ended work where hourly is the honest model.
The headline question is not 'which is cheaper' — over a well-run project the total can land in the same place. The real question is who carries the risk when the estimate is wrong, and how aligned your incentives are with the supplier's.
Under time-and-materials (T&M) you pay for every hour worked. If the build takes 40% longer than hoped, you pay 40% more. That is fair when the work is genuinely exploratory and no one could have scoped it honestly up front — but it puts the overrun risk squarely on you, and it quietly rewards a slower pace.
Under fixed-price milestones you agree a price for a defined deliverable. If it takes longer, that is the supplier's problem, not your budget's. The trade is that scope has to be pinned down first, and genuine changes mean a re-quote. For a business that needs to know the number before the board signs off, that certainty is usually worth more than the flexibility it gives up. See our breakdown of what an AI proof of concept costs for how a fixed PoC price is built.
Read this by the row that worries you most. If it is budget certainty or overrun risk, fixed-price wins. If it is a research question no one can scope yet, hourly is the honest answer.
| Dimension | Time & materials (hourly) | Fixed-price milestones |
|---|---|---|
| Budget certainty | Low — final cost depends on hours logged | High — the number is agreed before work starts |
| Scope risk | You absorb it; loose scope inflates the bill | Supplier absorbs it; scope is pinned, changes re-quoted |
| Who carries overrun risk | You (the client) | The supplier |
| Incentive alignment | Rewards more hours, not faster delivery | Rewards shipping the deliverable efficiently |
| Best-fit project type | Open-ended research, R&D, evolving discovery | Defined SME builds: audits, PoCs, production launches |
We will be straight with you: not everything should be fixed-price. If the honest answer to 'what does done look like?' is 'we won't know until we've experimented', a fixed price is either padded with a big risk premium or it will trap both sides into shipping the wrong thing to hit the milestone. In that case, T&M is the fairer model — you fund a genuine investigation and stop when you have your answer.
Typical hourly-fits: exploratory data science where the feasibility itself is the question, ongoing advisory or a fractional-CTO arrangement, and small iterative tweaks after launch. That is exactly why we keep consulting available at €150/hour alongside the fixed ladder — so the open-ended work is billed honestly by the hour rather than dressed up as a fixed scope.
We de-risk the parts that can be scoped and bill by the hour for the parts that can't. The ladder runs (all excluding VAT):
AI Audit & Strategy — €2,500. Fixed. We map where AI genuinely fits, size the value, and hand you a plan. The scope is clear, so the price is clear.
Proof of Concept — €20,000. Fixed. A working prototype on your own data, against success criteria we agree first. If it works, you know before spending production money.
Production — from €50,000. Fixed per milestone, scoped from what the PoC proved — and you own the code, not a demo. See why we ship production AI, not demos.
The audit-first sequence is the point: each fixed price is quoted on the back of what the previous stage actually revealed, so you are never asked to fix-price something nobody has scoped. Where the work is genuinely open-ended, it runs at €150/hour instead.
Fixed-price milestones with a senior boutique are not the right answer for every buyer. If you need dozens of engineers on a multi-year enterprise programme, a large consultancy's scale and bench fit better — our boutique vs Big Four comparison lays out honestly where each wins. If you already have a strong in-house AI team and just need extra hands, T&M contractors are usually cheaper than a fixed-scope build. And if cost is the single dominant factor, an offshore team will quote lower per hour than we will — that is a real advantage we won't pretend away. We are the right fit when you want senior people, a fixed number you can plan around, and to own what gets built.
No. Over a cleanly-run project the total cost can land in roughly the same place. Fixed-price does not necessarily save money — it moves the overrun risk from you to the supplier and gives you a number you can plan around. If a build runs long under T&M you pay the difference; under a fixed price the supplier absorbs it. You are buying certainty, not automatically a discount.
When the scope is genuinely open-ended and no one can honestly define 'done' yet — exploratory research, feasibility questions where the answer might be 'no', ongoing advisory, or small iterative changes after launch. Forcing a fixed price onto that kind of work either loads it with a risk premium or pushes both sides to ship the wrong thing to hit a milestone. That is why we bill open-ended work at €150/hour.
Because most SME AI work can be scoped once you do the groundwork, and a business planning a budget deserves to know the number up front. Our ladder — €2,500 audit, €20,000 PoC, production from €50,000 — quotes each fixed stage on the back of what the previous one revealed, so we never fix-price something that hasn't been scoped. The audit de-risks the estimate before you commit to a build.
A clearly scoped deliverable with agreed success criteria written down before work starts, plus a simple change process: genuine additions to scope are re-quoted, not absorbed silently or argued over. Because the price is fixed, our incentive is to ship the agreed thing efficiently rather than to run up hours — the interests line up. The upfront PoC is where most scope ambiguity gets removed.
Yes, and it is often the right structure. We fix-price the parts that can be scoped — audit, PoC, defined production milestones — and bill genuinely open-ended work (advisory, research spikes, post-launch tweaks) at €150/hour. That keeps the certainty where it matters and the honesty where it matters, rather than pretending exploratory work is a fixed deliverable.
Often, yes — and we won't pretend otherwise. A freelancer or offshore team will usually quote a lower hourly rate. The trade-offs are who carries the overrun risk, how senior the hands on your project are, and whether you end up owning production-grade code you can maintain. If low cost per hour is the dominant factor, they may fit better. We compete on senior delivery, a fixed number you can plan around, and full ownership — see our honest look at AI consultants in the Netherlands.
Tell us what you're trying to build and we'll tell you honestly whether it should be fixed-price or hourly — in a free 30-minute consultation, before any quote.
Book a free consultation →